11 karma · joined June 8, 2026
That to me seems to address a lot of the IO facing pitfalls formal verification can struggle with. At the obvious cost of it being a bear to implement.
I’m still pretty early on in the explore phase. Once I get through a cleanup pass or ten I’ll see if I feel good enough about it to share haha.
Overall it’s been a fun learning experience and I’m looking forward to some more of the hardware work I’ll need to jump into soon. I really want to get a more focused kitchen / cooking oriented voice assistant working. So far I have a few simple voice-to-timer settings done e.g. “set a 10 minute timer for the pasta” that tells me “ding! Pasta timer” when it goes off. You can set as many concurrent timers as you need with different names.
I need some better hardware before I try using the pi for full hands free while cooking. I’ve mostly been using a webapp on my phone but afaik you can’t easily wake word a phone on a web app without some real hacking.
Overall the projects been enjoyable, once you understand the basics of a harness it feels like there’s a lot of problems you can throw them at.
For voice related things you have a lot of turn of phrase scenarios that can make no sense unless you know. Phrasing like “Put Larry on the horn.” For someone familiar with old lingo for phone calls makes sense. For someone else they might think of a war horn, someone else a music class.
All of those are wildly different situations. It’s not hard to see how one oops between two non deterministic things can quickly go off the rails.
The fact we can get away with so much non-determinism->non-determinism recursion is frankly amazing when you realize how easy it is to imprecisely describe what it is you’re thinking.
I’m working on a tiny agent harness at home to learn and the process of taking human speech and turning it into agent tool calls that output something generally deterministic depending on how the tool is defined is so interesting.
One of the big takeaways is you really only have to rely on the non-determinism<->determinism translation layer once when you switch between the two domains. You can obviously rely on it more if you want, and that’s probably faster because determinism is hard, but you don’t need too do that.
I failed to realize that I needed to secure a studentaid .gov account someone was able to open in my name with data breach information.
Thankfully my credit was frozen so I didn’t need to untangle an actual loan, but it would have been a huge legal mess otherwise.
I guess my fear is what account am I going to miss securing next that leads to a giant life ruining problem? If I didn’t setup credit freezes someone else could have with the info in the breaches. I didn’t even think to secure a studentaid account I didn’t know existed. In theory having those credit bureau accounts frozen should be enough, but anyone with enough information on you can likely recover them regardless.
To me the whole experience really drives home how much of a joke the security on a lot of this is. Anyone who seriously sets their eyes on you can just totally ruin your life if they’re dedicated enough.
Though most people doing this its much more effective to take advantage of people who don’t know any better. Credit not frozen, loan accounts not made or secured, etc. Pwning 20 people doing nothing will always be better ROI than trying to PWN one person with their stuff in order. Until you piss the wrong person or they think you’re worth the effort.
I guess I can see how you can view it as not your problem. But there are only so many grandmas to scam. The whole problem space to me metaphorically is everyone’s door is wide open, grandmas is just a straight shot to get in. Mine? Well I have some ball bearings, calipers, and a moat but the doors still open. It’s not like someone is going to rob my open house instead of grandma. I only have to dodge all the traps when I leave and come back but that’s whatever.
The whole thing is absurd. We have doors and locks and better ways to do this and instead we just live like this?